Academia.eduAcademia.edu

Figure 7 – uploaded by 宸琮 邢

See full PDFdownloadDownload figure
Fig. 6. Precision@k of link prediction for different data sets (dimension of embedding is 128). Fig. 7. MAP of link prediction for different data sets with varying dimensions.

Figure 6 Precision@k of link prediction for different data sets (dimension of embedding is 128). Fig. 7. MAP of link prediction for different data sets with varying dimensions.

Related Figures (12)

List of graph embedding approaches
List of graph embedding approaches
3.5.1 SDNE
3.5.1 SDNE
Fig. 1. Examples illustrating the effect of type of similarity preserved. Here, CPE and SPE stand for Community Preserving Embedding anc Structural-equivalence Preserving Embedding, respectively.
Fig. 1. Examples illustrating the effect of type of similarity preserved. Here, CPE and SPE stand for Community Preserving Embedding anc Structural-equivalence Preserving Embedding, respectively.
6 EXPERIMENTS AND ANALYSIS HEP-TH [63]: The original dataset contains abstracts of papers in High Energy Physics Theory for the period from Jan- uary 1993 to April 2003. We create a collaboration network for the papers published in this period. The network has 7,980 nodes and 21,036 edges.
6 EXPERIMENTS AND ANALYSIS HEP-TH [63]: The original dataset contains abstracts of papers in High Energy Physics Theory for the period from Jan- uary 1993 to April 2003. We create a collaboration network for the papers published in this period. The network has 7,980 nodes and 21,036 edges.
Fig. 2. Precision@k of graph reconstruction for different data sets (dimension of embedding is 128).
Fig. 2. Precision@k of graph reconstruction for different data sets (dimension of embedding is 128).
Fig. 3. MAP of graph reconstruction for different data sets with varying dimensions 6.2 Visualization LE an
Fig. 3. MAP of graph reconstruction for different data sets with varying dimensions 6.2 Visualization LE an
Fig. 4. Visualization of SBM using t-SNE (original dimension of embedding is 128). Each point corresponds to a node in the graph. Color of a nod denotes its community.
Fig. 4. Visualization of SBM using t-SNE (original dimension of embedding is 128). Each point corresponds to a node in the graph. Color of a nod denotes its community.
g. 5. Visualization of Karate club graph. Each point corresponds to a nox onerated by HOPE and SDNE which preserve higher order roximities well separate the communities although as the ata is well structured LE, GF and LLE are able to capture ymmunity structure to some extent.
g. 5. Visualization of Karate club graph. Each point corresponds to a nox onerated by HOPE and SDNE which preserve higher order roximities well separate the communities although as the ata is well structured LE, GF and LLE are able to capture ymmunity structure to some extent.
Fig. 8. Micro-F1 and Macro-F1 of node classification for different data set: focusing on interpreting the embedding learned by these models can be very fruitful. Utilizing embedding to study graph evolution is a new research area which needs further exploration. Recent work by [66] and [67] pursued this line of thought and illustrate how embeddings can be used for dynamic graphs. Generating synthetic networks with real- world characteristics has been a popular field of research [68] primarily for ease of simulations. Low dimensional vector representation of real graphs can help understand their structure and thus be useful to generate synthetic graphs with real world characteristics. Learning embedding with a generative model can help us in this regard. The authors are supported by DARPA (grant number D16AP00115), [ARPA (contract number 2016-16041100002), and AFRL (contract number FA8750-16-C-0112). The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of DARPA, IARPA, AFRL, or the U.S. Government. The U.S. Govern- ment had no role in study design, data collection and anal- ysis, decision to publish, or preparation of the manuscript. The U.S. Government is authorized to reproduce and dis- tribute reprints for governmental purposes notwithstanding any copyright annotation therein.
Fig. 8. Micro-F1 and Macro-F1 of node classification for different data set: focusing on interpreting the embedding learned by these models can be very fruitful. Utilizing embedding to study graph evolution is a new research area which needs further exploration. Recent work by [66] and [67] pursued this line of thought and illustrate how embeddings can be used for dynamic graphs. Generating synthetic networks with real- world characteristics has been a popular field of research [68] primarily for ease of simulations. Low dimensional vector representation of real graphs can help understand their structure and thus be useful to generate synthetic graphs with real world characteristics. Learning embedding with a generative model can help us in this regard. The authors are supported by DARPA (grant number D16AP00115), [ARPA (contract number 2016-16041100002), and AFRL (contract number FA8750-16-C-0112). The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of DARPA, IARPA, AFRL, or the U.S. Government. The U.S. Govern- ment had no role in study design, data collection and anal- ysis, decision to publish, or preparation of the manuscript. The U.S. Government is authorized to reproduce and dis- tribute reprints for governmental purposes notwithstanding any copyright annotation therein.
Fig. 9. Micro-F1 and Macro-F1 of node classification for different data sets varying the number of dimensions. The train-test split is 50%. REFERENCES [17] N. Friedman, L. Getoor, D. Koller, and A. Pfeffer, “
Fig. 9. Micro-F1 and Macro-F1 of node classification for different data sets varying the number of dimensions. The train-test split is 50%. REFERENCES [17] N. Friedman, L. Getoor, D. Koller, and A. Pfeffer, “
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved